用pandas函数代替下列语句：def RMSE_MAPE(ydata,y_hatdata,indexlist): y=[] y_hat=[] for i in indexlist: y.append(ydata.iat[i,0]) y.append(ydata.iat[i,1]) y.append(ydata.iat[i,2]) y_hat.append(y_hatdata.iat[i,0]) y_hat.append(y_hatdata.iat[i,1]) y_hat.append(y_hatdata.iat[i,2]) x1=0 for i in range(len(indexlist)): x1=x1+np.square(y[i] - y_hat[i]) x2=0 for i in range(len(indexlist)): x2=x2+np.abs((y[i] - y_hat[i]) / y[i]) RMSE = np.sqrt(x1/len(indexlist)) MAPE = x2*100/len(indexlist) print('RMSE:'+str(RMSE)+' MAPE:'+str(MAPE)) return

import pandas as pd

def RMSE_MAPE(ydata, y_hatdata, indexlist):
y = pd.concat([ydata.loc[indexlist, 'col1'], ydata.loc[indexlist, 'col2'], ydata.loc[indexlist, 'col3']], ignore_index=True)
y_hat = pd.concat([y_hatdata.loc[indexlist, 'col1'], y_hatdata.loc[indexlist, 'col2'], y_hatdata.loc[indexlist, 'col3']], ignore_index=True)

x1 = np.square(y - y_hat).sum()
x2 = np.abs((y - y_hat) / y).sum()

RMSE = np.sqrt(x1 / len(y))
MAPE = x2 * 100 / len(y)

print('RMSE:', RMSE, 'MAPE:', MAPE)
return

相关问题

用pandas函数代替下列语句：def RMSE_MAPE(ydata,y_hatdata,indexlist): y=[] y_hat=[] for i in indexlist: y.append(ydata.iat[i,0]) y.append(ydata.iat[i,1]) y.append(ydata.iat[i,2]) y_hat.append(y_hatdata.iat[i,0]) y_hat.append(y_hatdata.iat[i,1]) y_hat.append(y_hatdata.iat[i,2]) x1=0 for i in range(len(indexlist)): x1=x1+np.square(y[i] - y_hat[i]) x2=0 for i in range(len(indexlist)): x2=x2+np.abs((y[i] - y_hat[i]) / y[i]) RMSE = np.sqrt(x1/len(indexlist)) MAPE = x2*100/len(indexlist) print('RMSE:'+str(RMSE)+' MAPE:'+str(MAPE)) return

可以使用pandas中的iloc函数简化代码：

def RMSE_MAPE(ydata, y_hatdata, indexlist):
y = ydata.iloc[indexlist, :].values.flatten()
y_hat = y_hatdata.iloc[indexlist, :].values.flatten()
x1 = np.square(y - y_hat).sum()
x2 = np.abs((y - y_hat) / y).sum()
RMSE = np.sqrt(x1/len(indexlist))
MAPE = x2*100/len(indexlist)
print(f'RMSE: {RMSE:.4f} MAPE: {MAPE:.2f}')
return

用pandas函数简化下列语句：def RMSE_MAPE(ydata,y_hatdata,indexlist): y=[] y_hat=[] for i in indexlist: y.append(ydata.iat[i,0]) y.append(ydata.iat[i,1]) y.append(ydata.iat[i,2]) y_hat.append(y_hatdata.iat[i,0]) y_hat.append(y_hatdata.iat[i,1]) y_hat.append(y_hatdata.iat[i,2]) x1=0 for i in range(len(indexlist)): x1=x1+np.square(y[i] - y_hat[i]) x2=0 for i in range(len(indexlist)): x2=x2+np.abs((y[i] - y_hat[i]) / y[i]) RMSE = np.sqrt(x1/len(indexlist)) MAPE = x2*100/len(indexlist) print('RMSE:'+str(RMSE)+' MAPE:'+str(MAPE)) return

可以用 pandas 的 loc 函数代替循环中的 iat 函数，并利用矩阵转置函数和 numpy 中的 power 和 mean 函数简化代码，具体如下：

import numpy as np
import pandas as pd

def RMSE_MAPE(ydata, y_hatdata, indexlist):
y = ydata.loc[indexlist].values.T
y_hat = y_hatdata.loc[indexlist].values.T
x1 = np.power(y - y_hat, 2).mean()
x2 = np.abs((y - y_hat) / y).mean()
RMSE = np.sqrt(x1)
MAPE = x2 * 100
print(f'RMSE:{RMSE} MAPE:{MAPE}')
return RMSE, MAPE

其中，y 和 y_hat 利用 pandas 的 loc 函数获取 indexlist 对应的行数据，并利用 values 属性得到其 numpy 数组表示，同时利用矩阵转置函数将 y 和 y_hat 的列向量转置成行向量。x1 和 x2 利用 numpy 中的 power 和 mean 函数分别计算平方误差和绝对误差，并根据公式计算出 RMSE 和 MAPE。最后，利用 f-string 来输出 RMSE 和 MAPE 的值。